Phylogeography and Ecological Niche Modeling Unravel the Evolutionary

Phylogeography and ecological niche modeling unravel the evolutionary history of the African green toad, Bufotes boulengeri boulengeri (Amphibia: Bufonidae), through the Quaternary Violaine Nicolas, Abderrahmane Mataame, Pierre-André Crochet, Philippe Geniez, Soumia Fahd, Annemarie Ohler

To cite this version:

Violaine Nicolas, Abderrahmane Mataame, Pierre-André Crochet, Philippe Geniez, Soumia Fahd, et al.. Phylogeography and ecological niche modeling unravel the evolutionary history of the African green toad, Bufotes boulengeri boulengeri (Amphibia: Bufonidae), through the Quaternary. Journal of Zo- ological Systematics and Evolutionary Research, Wiley, 2018, 56 (1), pp.102-116. 10.1111/jzs.12185. hal-01724136

HAL Id: hal-01724136 https://hal.sorbonne-universite.fr/hal-01724136 Submitted on 6 Mar 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. 1 Phylogeography and ecological niche modelling unravel the evolutionary history of the 2 African green toad, Bufotes boulengeri boulengeri (Amphibia: Bufonidae), through the 3 Quaternary 4 5 Violaine NICOLAS1, Abderrahmane MATAAME2, Pierre-André CROCHET3, Philippe 6 GENIEZ4, Soumia FAHD5, Annemarie OHLER1 7 8 1 Institut de Systématique, Evolution, Biodiversité, ISYEB - UMR 7205 - CNRS, MNHN, 9 UPMC, EPHE, Muséum national d’Histoire naturelle, Sorbonne Universités, Paris, France. 10 [email protected], [email protected] 11 2 Institut Scientifique de Rabat, Rabat, Maroc. [email protected] 12 3 CEFE UMR 5175, CNRS, Université de Montpellier, Université Paul-Valéry Montpellier, 13 EPHE, Montpellier, France. [email protected] 14 4 PSL Research University, CEFE UMR 5175, CNRS, Université de Montpellier, Université 15 Paul-Valéry Montpellier, EPHE, Biogéographie et Ecologie des Vertébrés, Montpellier, 16 France. [email protected] 17 5 Laboratoire "Ecologie, Biodiversité et Environnement", Département de Biologie, Faculté 18 des Sciences de Tétouan, Université Abdelmalek Essaâdi, Tétouan, Maroc. 19 [email protected] 20 21 Corresponding author: Violaine NICOLAS 22 Tel: 00 33 1 40 79 48 52 23 [email protected] 24 25 Running title: Evolutionary history of the African green toad 26 27 Keywords 28 Amphibian - Ecological niche modelling - Genetic structure - Maghreb – Pleistocene 29

1 Abstract 2 Recent integration of ecological niche models in phylogeographic studies is improving our 3 understanding of the processes structuring genetic variation across landscapes. Previous 4 studies on the amphibian B. b. boulengeri uncovered a surprisingly weak intraspecific 5 differentiation across the Maghreb region. We widely sampled this species from Morocco to 6 Egypt, and sequenced one nuclear and three mitochondrial (mtDNA) genes to determine the 7 level of genetic variability across its geographic range. We evaluated these data with 8 ecological niche modelling to reveal its evolutionary history in response to climate change 9 during the Quaternary. Our results highlight some mtDNA phylogeographic structure within 10 this species, with one haplogroup endemic to coastal Morocco, and one haplogroup widely 11 distributed throughout North Africa. No or little genetic differentiation is observed between 12 isolated populations from the Hoggar Mountains, the Sabha district and the islands of 13 Kerkennah and Lampedusa, compared to others populations. This can be explained by the 14 expansion of the distribution range of B. b. boulengeri during glacial periods. This might have 15 facilitated the species’ dispersal and subsequent gene ﬂow between most North African 16 localities. 17 18 Introduction 19 Green toads of the Bufotes viridis (Laurenti, 1768) species complex are widespread in the 20 Palearctic region where they have differentiated into several lineages (Batista et al. 2006; 21 Stöck et al. 2006; Stöck et al. 2008). Using two mitochondrial genes, Stöck et al. (2006) 22 suggested that at least five species can be recognised within this complex of species, but 23 evidence of reproductive isolation or even distinct nuclear gene pools is lacking for several of 24 these mitochondrial lineages. More detailed studies using mtDNA and two nuclear sequence 25 markers, sex-chromosomes, and additional morphological, phenological, and bioacoustic data 26 focused on the North African-Sicilian lineage (Stöck et al. 2008) demonstrated its 27 evolutionary distinctiveness and documented strong reproductive isolation from the Italian 28 mitochondrial lineage in Sicily. On the contrary, studies of contact zones between the younger 29 lineages viridis and balearicus (Boettger, 1880) revealed extensive introgression of nuclear 30 markers and lack of intrinsic reproductive isolation (Dufresnes et al. 2014). As a consequence 31 Speybroeck et al. (2016) recognised only the North African-Sicilian lineage as a valid species, 32 Bufotes boulengeri (Lataste, 1879), and proposed to postpone further species-level splits until 33 the situation in other contact zones is documented. According to these authors B. boulengeri

1 is subdivided in two subspecies: B. b. boulengeri and B. b. siculus (Stöck et al. 2008) 2 (availability of the nomen siculus is unclear, see Dubois et al. 2013; ICZN 2014). 3 The African green toad, Bufotes b. boulengeri, occurs as fragmented populations in North 4 Africa from the Moroccan Atlantic coast eastwards to the Nile valley in Egypt, including 5 several Mediterranean islands (Martínez-Solano et al. 2015). It is a characteristic inhabitant of 6 relatively arid, open landscapes (Beukema et al. 2013) but it can inhabit a wide range of 7 habitat: forested areas, scrubland, dry grassland, semi-deserts and deserts (Martínez-Solano et 8 al. 2015). Several highly isolated populations are known from Algeria (e.g. Hoggar 9 Mountains), Libya (Sabha district), as well as on the islands of Kerkennah and Lampedusa 10 (Figure 1). In Morocco it is most abundant on the Atlantic coast (as far south as Dakhla) and 11 in the High and Middle Atlas Mountains, with scattered populations in the north, north-east 12 and east of the country and in the northern Saharan margins (Beukema et al. 2013). The 13 species is mostly absent from the Rif Mountains, where it is limited to river corridors and is 14 absent from the mountain cores characterized by a more humid climate (Donaire et al. 2011). 15 While strong phylogeographic structure is commonly found in amphibian species across the 16 Maghreb region (Ben Hassine et al. 2016; Nicolas et al. 2015; Recuero et al. 2007; Vences et 17 al. 2014), previous studies on B. b. boulengeri uncovered a surprisingly weak intraspecific 18 differentiation over its geographic range (Batista et al. 2006; Stöck et al. 2006; Stöck et al. 19 2008). This is especially unexpected because fossil data indicates that the African green toad 20 has been present in North Africa since the beginning of the Pleistocene (2.5 Ma in Morocco, 21 (Bailon 2000). Climatic reconstructions and paleoenvironmental data show that North Africa 22 regularly shifted from wetter to drier climatic conditions throughout the Quaternary 23 (deMenocal 2004; Schuster et al. 2006; Tabel et al. 2016). This climatic instability led to 24 periodic modifications of habitats, such as contraction/ expansion of the Sahara desert, 25 emergence of more open habitats or deep modifications of hydrographic networks (Rognon 26 1989; Le Houérou 1997; Drake et al. 2011). Several geographical features, such as the Sahara 27 desert, the Atlas Mountains or the Moulouya River valley are known to have acted as barrier 28 to dispersal for many species and promoted diversification in several species or species 29 complexes ( Aulagnier and Thevenot 1986; Alvarez et al. 2000; Brown et al. 2002; Ben Faleh 30 et al. 2012; Douady et al. 2003; Escoriza et al. 2006; Fritz et al. 2006; Barata et al. 2008; 31 Nicolas et al. 2014). 32 In this study, we widely sampled African green toads from Morocco to Egypt (95 newly 33 collected individuals in Morocco, Algeria and Egypt, and 39 specimens from Algeria, Egypt, 34 Libya, Morocco, Tunisia and Lampedusa Island retrieved from GenBank) and sequenced

1 three mitochondrial (mtDNA) and one nuclear (nDNA) genes to determine the level of 2 genetic variability within B. b. boulengeri across its geographic range. We specifically tested 3 whether isolated populations from the Hoggar Mountains, the Sabha district and the islands of 4 Kerkennah and Lampedusa are genetically differentiated from others populations. Moreover, 5 we evaluated these genetic data with ecological niche modelling based on bioclimatic data to 6 reveal the evolutionary history of B. b. boulengeri in response to climate change during the 7 Quaternary. 8 9 Material and methods 10 DNA extraction, amplification and sequencing 11 Ninety-five newly collected individuals were included in this study (15 specimens from 12 Algeria, two from Egypt and 78 from Morocco; Supporting information Table S1). When 13 available, these specimens were deposited either at the Museum National d’Histoire Naturelle 14 (MNHN, Paris, France) or at the collection of the UMR5175 CEFE-EPHE (Collection BEV, 15 Montpellier, France); for several specimens only tissues samples were available. Genomic 16 DNA was extracted from liver, spleen, muscle or toes preserved in 95% ethanol using the 17 NucleoSpin Tissue Core kit (Macherey-Nagel, Hoerdt, France). The mitochondrial D-loop 18 region (D-loop), 16S rRNA gene (16S; mtDNA), cytochrome c oxidase subunit 1 gene (CO1; 19 mtDNA) and recombination activating protein 1 gene (RAG1; nDNA) were amplified using 20 polymerase chain reaction (PCR) primers ControlB-H and CytbA-L (Goebel et al. 1999), 21 16Sar-L and 16Sbr-H (Palumbi et al. 1991), Amp-P3F and Amp-P3R (San Mauro et al. 2004) 22 and MartFL1 and AmpRI (Chiari et al. 2004), respectively (see Supporting information Table 23 S2 for primer sequences). The double-stranded PCR products were purified and sequenced at 24 the Genoscope (Ivry-sur-Seine, France) or at EUROFINS (Ebersberg, France) using the 25 Sanger dideoxy sequencing run on an ABI 3730XL sequencer (Applied Biosystems, Foster 26 City, CA, USA). All newly determined haplotypes were submitted to GenBank (accession 27 numbers KY498904 to KY499005). 28 In our analyses, we also included 39 specimens of B. b. boulengeri available in the GenBank 29 database for which the D-loop, 16S and/or the CO1 genes were sequenced. 30 The combination of these two datasets (newly sequenced individuals and Genbank data) 31 allowed us to include specimens from six countries and 40 localities, covering most of the 32 geographical range of the species (Figure 1). A total of 122, 113, 88 and 35 individuals were 33 included in our analyses for the D-loop, 16S, CO1 and RAG1 DNA regions, respectively 34 (Table 1).

1 2 Median-joining network analyses 3 Relationships between haplotypes were inferred by constructing a network using the median- 4 joining method available in NETWORK v4.500 (Bandelt et al. 1999). This method accounts 5 for the coexistence of ancestral and descendent haplotypes, multifurcations and reticulate 6 relationships (Posada and Crandall 2001b). We used the maximum parsimony post-processing 7 option, which removes all superfluous median vectors and links that are not contained in the 8 shortest trees of the network. Sequences of 545, 586, 806 and 397 bp were retained for the 9 haplotype network analyses, for the D-loop, 16S, CO1 and RAG1 genes, respectively, to 10 minimize the number of incomplete sequences; as adding ambiguous data in median-joining 11 trees is problematic. For the same reason, we treated all genes separately in haplotype 12 network analyses as sequences of one or two genes were missing for several specimens (Table 13 1 and Supporting information Table S1). 14 Alignment of 16S gene sequences required the insertion of three gaps of length one, and 15 alignment of D-loop sequences required the insertion of 11 gaps: eight of length one and one 16 of length three. Indels were taken into account in phylogenetic analyses, but were removed 17 from genetic diversity and demographic history analyses. 18 For the nuclear RAG1 gene, the existence of heterozygous positions was investigated and an 19 input file constructed from this information using SeqPHASE (Flot 2010). The phase of each 20 haplotype and its reconstruction were carried out using PHASE v.2.1.1 ( Stephens et al. 2001; 21 Stephens and Scheet 2005) by running the formerly built input file and by considering the 22 default parameters of the software. Phased haplotypes were used in all subsequent analyses. 23 The phasing of all specimens was straightforward (probability of 1.0), except for three of 24 them (probability below 0.6). These three specimens were therefore removed from subsequent 25 analyses (Supporting information Table S1). 26 27 Genetic diversity, population differentiation and demographic history 28 The average number of nucleotide differences, nucleotide diversity and haplotype diversity 29 (Nei 1987) were calculated using DNASP 5.10 (Librado and Rozas 2009). 30 Inferences of population expansion were made using two different methods. First we 31 calculated the Fs statistic of Fu (Fu 1997), which is based on the probability of having a 32 number of haplotypes greater or equal to the observed number of samples drawn from a 33 constant-sized population. Secondly we calculated the R2 statistic (Ramos-Onsins and Rozas 34 2002), which is based on the difference between the number of singleton mutations and the

1 average number of nucleotide differences. Ramos-Onsins and Rozas (2002) demonstrated 2 that these statistics have the greatest power to detect population expansion for non- 3 recombining regions of the genome under a variety of different circumstances, especially 4 when population sample sizes are large or when sample sizes are small. The significance of 5 Fu’s Fs and R2 was obtained by examining the null distribution of 1,000 coalescent 6 simulations of these statistics using DNASP 5.10. Significantly large negative Fu’s 7 Fs values and significantly positive R2 values were taken as evidence of a population 8 expansion. The Fs statistic was considered significant when the p-value was below 0.02. 9 We used Spatial analysis of Shared Alleles (SaShA) to test whether there is any genetic 10 subdivision across the geographical space. This method can detect subtle genetic 11 differentiation even when diversity is moderate, gene flow is moderate or high, and when 12 sampling is poor –that is low sample size, or samples of uneven sizes and taken at uneven 13 spatial intervals (Kelly et al. 2010). This analysis uses spatial and allele (haplotype) 14 information to detect non-random allele distribution against an expectation of panmixia. The 15 test statistic OM describes the observed mean distance between alleles. Where OM is less 16 than the expected mean (EM), alleles are considered to be aggregated (underdistributed). 17 Where OM is larger than EM, alleles are considered overdispersed (panmictic). A jackknifing 18 procedure allows the identification of which alleles contribute to the distribution. 19 20 Phylogenetic analysis 21 To better understand phylogenetic relationships between the two haplogroups identified in the 22 median-joining network analyses we also performed phylogenetic reconstruction on the 16S 23 and D-loop sequences combined using Bayesian Inference (BI). Only specimens for which 24 the two genes were available were included in this analysis. CO1 sequences were not included 25 because we had no data for Tunisia and Egypt. We used MrBayes v. 3.2.5 (Ronquist & 26 Huelsenbeck 2003). Analyses were conducted with two runs and four chains (three hot and 27 one cold), each of 5,000,000 generations, with trees sampled every 1,000 generations. 28 Stationarity was assessed by examining the average SD of split frequencies and the potential 29 scale reduction factor. After a visual inspection, the first 25% of iterations was excluded as 30 burn-in time and the resulting trees were combined in a majority rule consensus tree to obtain 31 posterior probabilities. We used a two partition model (one for each gene). Each gene data set 32 was tested for the most appropriate model of sequence evolution using the Akaike 33 information criterion (Akaike 1973) test as implemented in MrModeltest v. 3.7 (Posada & 34 Crandall 2001a) using MrMTgui (available from http://genedrift.org/software/mrmtgui.html).

1 The HKY + I + G and HKY + I models were chosen for the D-loop and 16S, respectively. To 2 root our tree, we included four sequences of B. b. siculus (GenBank EU497543, EU497553, 3 EU497538 and EU497548 for D-loop; EU497443, EU497453, EU497438 and EU497448 for 4 16S), the sister clade of B. b. boulengeri, as well as three most distantly related outgroups: B. 5 viridis (EU497516 and EU497416), B. balearicus (EU497577 and EU497477) and B. 6 variabilis (FJ882812 and DQ629623) (Dufresnes et al. 2014; Stöck et al. 2008). 7 8 Divergence time estimates 9 Obtaining robust divergence time estimates is not a trivial task (Ho et al. 2014). In the 10 absence of a good dated fossil record for the genus Bufotes we used two approaches to 11 estimate divergence times: we used the substitution rates derived from the paper of Stöck et 12 al. (2008) and a secondary calibration point derived from the work of Garcia-Porta et al. 13 (2012). 14 In these analyses, we only took into account the D-loop and 16S results as they are the only 15 genes for which background data on substitution rates and DNA sequences of sister species 16 were available. Divergence time estimates were inferred using BEAST version 1.4.7 17 (Drummond and Rambaut 2007). We used an uncorrelated lognormal Bayesian relaxed clock. 18 To account for the fact that our trees mostly describe intraspecific relationships, we used a 19 coalescent model tree prior with a Bayesian skyline model of population size. The computer 20 program MRMODELTEST (Nylander 2004) was used to evaluate the fit of 24 nested models 21 of nucleotide substitution to the data. The model chosen by MRMODELTEST according to 22 the Akaike information criterion was then used in BEAST. Two distinct runs were carried out 23 for each gene separately, each one with four independent chains of 60,000,000 generations, 24 default priors, and trees sampled every 6,000 generations. Results were visually inspected 25 using TRACER version 1.5 to ensure proper mixing of the Markov chain Monte Carlo. We 26 applied a conservative burn-in of 25%. Samples from both runs were combined using the 27 software LOGCOMBINER version 1.4.7, and a consensus chronogram was obtained using 28 the program TREEANNOTATOR version 1.6.2 using the options maximum clade credibility 29 tree and median heights. To estimate divergence times we used to approaches. First, we 30 calibrated our tree using the same range of substitution rates as Stöck et al. (2008), which 31 were derived from several general works on amphibians: we specified the prior for the mean 32 substitution rate as a normal distribution, with a mean of 0.02 and a standard deviation of 33 0.007 substitution per site per My for the D-loop, and a mean of 0.008 and a standard 34 deviation of 0.003 substitution per site per My for the 16S gene. To root our trees, we

1 included four sequences of B. b. siculus (GenBank EU497543, EU497553, EU497538 and 2 EU497548 for D-loop; EU497443, EU497453, EU497438 and EU497448 for 16S), the sister 3 clade of B. b. boulengeri (Dufresnes et al. 2014; Stöck et al. 2008). The model chosen by 4 MRMODELTEST was HKY+I+G for the D-loop region, and HKY+I for the 16S gene. 5 Second, we used a secondary calibration point derived from the work of Garcia-Porta et al. 6 (2012). These authors estimated the separation between the Iberian and the African 7 populations of Bufo bufo (Linnaeus, 1758) at 3.07 Ma (95% HPD = 1.91-4.36 Ma), based on 8 four calibration points derived from geological and fossil evidences. We used a normal prior 9 with a mean of 3.07 and a standard deviation of 0.70. The model chosen by 10 MRMODELTEST for the 16S gene was GTR+G. We included in these analyses all our 16S 11 sequences as well as sequences of B. bufo from Tunisia (Genbank JQ348746, JQ348749), 12 Morocco (JQ348755, JQ348750) and Iberia (JQ348689, JQ348710, JQ348725). 13 14 Species distribution modeling 15 We modeled the climatic niche of B. b. boulengeri to approximate its current, Last Glacial 16 Maximum (LGM) and mid-Holocene distributions. We applied ecological niche modeling 17 methods, where environmental data are extracted from current occurrence records, and habitat 18 suitability is evaluated across the landscape using program-specific algorithms (Elith et al. 19 2006). The present-day models were then projected on the climatic reconstructions of the 20 LGM (a cold and dry period, about 22,000 years ago ) and mid-Holocene (a wetter period, 21 about 6,000 years ago ) under the assumption that the climatic niche of the species remained 22 conserved through time (Elith et al. 2010). 23 Distribution records of African green toads were assembled from literature (Bons and Geniez 24 1996; Frynta et al. 2000; Baha El Din 2006; Stöck et al. 2006; Fahd and Mediani 2007; Harris 25 et al. 2008; Ibrahim 2008; Stöck et al. 2008; Brunet et al. 2009; Sicilia et al. 2009; García- 26 Muñoz et al. 2010; Amor et al. 2011; Donaire et al. 2011; Filippi et al. 2011; Beukema et al. 27 2013;Ibrahim 2013; Damas-Moreira et al. 2014; Márquez-Rodríguez 2014; Essghaier et al. 28 2015; Rosado et al. 2016), museum collections (Global Biodiversity Information Facility, 29 http://www.gbif.org, last access date: 26 August 2016) and our own unpublished databases 30 (Supporting information Table S3). To correct sampling bias, a subsample of records 31 regularly distributed in the geographical space was selected using ENMtools ver. 1.3 (Warren 32 et al. 2010). This method is the most efficient at correcting sampling bias (Fourcade et al. 33 2014). The resolution of the reference grid was set to 2.5 arc min (nearly 5 × 5 km), that is the 34 same as that used for the environmental layers. This coarse resolution was chosen to better

1 match the coordinate uncertainties associated with georeferenced, textual localities of 2 museum specimens. After filtering, the final dataset used for modelling consisted of 702 3 distribution records (Supporting information Figure S1). 4 As environmental layers, we used available climatic data from the WorldClim database 5 (Hijmans et al. 2005). We removed from our analyses variables BIO2, BIO3, BIO5, BIO7 and 6 BIO15 because those variables show a high level of discrepancy between General Circulation 7 Models (GCMs) for North Africa (Varela et al. 2015). Colinearity of the initial variables was 8 measured by Pearson’s correlation coefficient in ENMtools v1.3 (Warren et al. 2010). A total 9 of five variables, all of which had a Pearson correlation coefficient lower than 0.75, were 10 retained. The final set of environmental predictor variables used for the species distribution 11 models (SDM) consisted of: Annual Temperature (BIO1), Precipitation of Wettest Quarter 12 (BIO16), Precipitation of Driest Quarter (BIO17), Precipitation of Warmest Quarter (BIO18) 13 and Precipitation of Coldest Quarter (BIO19). 14 Climatic variables were used for present conditions and for the LGM and mid-Holocene. 15 Paleoclimate data were drawn from the general circulation model (GCM) simulations based 16 on two climate models: the Community Climate System Model (CCSM, version 3) (Collins et 17 al. 2006) and the Model for Interdisciplinary Research on Climate (MIROC, version 3.2) 18 (Hasumi and Emori 2004). For a discussion of the uncertainties associated with the climatic 19 data, see Schorr et al. (2012) and Varela et al. (2015). The use of these two different climate 20 models enabled us to assess and account for modeling uncertainty due to LGM climate data. 21 To predict the potential distribution of the species in current conditions and in the LGM, we 22 used MAXENT ver. 3.3.3 (Phillips et al. 2006), which returns a model with relative 23 occurrence probability of a species within the grid cells of the study area. To ensure 24 consistency of model predictions among repeated runs, we performed a 10-fold cross- 25 validation with random seed. Several values of the regularization multiplier were tested, as 26 suggested by Radosavljevic & Anderson (2014), and a final value of 2 was retained in our 27 final analyses. To determine whether the predictions for current conditions generated by 28 Maxent were better than random predictions, we used the area under the receiver operating 29 characteristic curve (AUC), a commonly used measurement for comparison of model 30 performance (Elith et al. 2006). The AUC ranges from 0 to 1, with greater scores indicating 31 better discrimination ability. An AUC greater than 0.5 indicates that the model discriminates 32 better than random. 33 34 Results

1 Molecular genetic analyses 2 The D-loop region is clearly the most variable of the four analyzed DNA regions, with a 3 higher number of polymorphic sites, a higher number of haplotypes, and higher haplotype and 4 nucleotide diversities (Table 2). The RAG1 nDNA fragment is the less variable one. 5 6 The D-loop network allows the identification of two main haplogroups (Figure 2). 7 Haplogroup 1 groups 44 specimens (10 haplotypes) coming from coastal Morocco, from 8 Aglou (locality 37) to Merja Zerga (locality 26; Figure 3). Haplogroup 2 is widely distributed 9 throughout North Africa and groups 78 specimens (24 haplotypes) coming from Morocco to 10 Egypt. Haplogroup 1 can be subdivided into two sub-haplogroups (1a and 1b) having 11 sympatric geographic distributions in Morocco (Supporting information Table S1). 12 Results of the 16S and CO1 networks are congruent with those obtained with the D-loop 13 network, and confirm the presence of one haplogroup restricted to Morocco, and one 14 haplogroup widely distributed throughout North Africa. No sub-structure is evident within the 15 Moroccan haplogroup. 16 The RAG1 gene recovers only 13 haplotypes, differing from one another by one to five 17 mutations. No phylogeographic structure is evident in the network: the three most common 18 haplotypes differ from one another by one to three mutations and are geographically widely 19 distributed. Haplotype h02 is present in the Algerian and Moroccan localities 17, 27, 28, 29, 20 34, 35, 38 and 39. Haplotype h01 is present in the Moroccan localities 27, 28, 29, 34 and 35, 21 and haplotype h03 are present in the Moroccan localities 28, 33, 35 and 38. 22 23 For the D-loop region, the SAShA analysis of all specimens reveals that B. boulengeri 24 haplotypes are significantly underdistributed (OM = 213 km, EM = 1285 km; p-value = p- 25 value < 0.001). This is also true for haplogroups 1a (OM = 125 km, EM = 185 km; p-value = 26 0.002) and 2 (OM = 250 km, EM = 1495 km; p-value < 0.001), but not for haplogroup 1b 27 which seems evenly distributed (OM = 208 km, EM = 201 km; p-value = 0.748). Our Jacknife 28 SAShA analyses indicate the robustness of the overall results, which remain qualitatively the 29 same when any one haplotype is removed. Within haplogroup 1a the most common haplotype 30 (16 individuals) is restricted to localities situated in-between Sidi Mokhtar (locality 35 on 31 Figure 1) and Berrechid (locality 28). Within haplogroup 2 the most common haplotype (22 32 individuals) is restricted to the Moroccan localities situated in-between the localities of Sidi 33 Mokhtar (locality 35), Merja Zerga (locality 26) and Al Baten (locality 25). Only one 34 haplotype is present in the Hoggar Mountains (10 individuals) and the Lampedusa Island (6

1 individuals), respectively, and each of these three haplotypes is endemic to each of these 2 regions. One haplotype is endemic to Kerkennah Islands (3 individuals). One haplotype (5 3 individuals) in found on the Mediterranean coast from Cap Bon (Tunisia, locality12) to 4 Sahahhat (Libya, locality 6). Finally one haplotype (3 individuals) is endemic to the three 5 southernmost Moroccan localities of Tantan (locality 40), Abtith (locality 38) and Saqiyat al- 6 Hamra (locality 39). In Sasha analyses only haplotypes found in at least three individuals are 7 considered. We would like to underline that one haplotype represented by two individuals in 8 our dataset was found in both Kekennah islands (locality 11) and Wadi El Natrum (locality 2). 9 For both the 16S and CO1 genes, SASha analyses of all specimens reveals that B. b. 10 boulengeri haplotypes are significantly underdistributed (OM = 266 km, EM = 996 km for 11 16S; OM = 264 km, EM = 756 km for CO1; p-values < 0.001). This is also true for 12 haplotypes of haplogroup 2 (OM = 352 km, EM = 1254 km for 16S; OM = 345 km, EM = 13 1099 km for CO1; p-values < 0.001); while haplotypes of haplogroup 1 are evenly distributed 14 (OM = 180 km, EM = 190 km for 16S; OM = 191 km, EM = 183 km for CO1; p-values = 15 0.221 and 0.433, respectively). 16 For the 16S gene, the most common haplotype within haplogroup 2 (23 individuals) is 17 restricted to the Moroccan localities situated in-between the localities of Abtith (locality 38), 18 Merja Zerga (locality 26) and Al Baten (locality 25). One haplotype (16 individuals) is 19 endemic to the Hoggar Mountains (localities 17 to 22) and Al Baten (locality 25), and it is the 20 only haplotype found in the Hoggar Mountains. Two haplotypes (8 and 5 individuals, 21 respectively) are present in both coastal Tunisia (localities 9 or 12) and Kerkennah Island 22 (locality 11). One haplotype (5 individuals) is endemic to Lampedusa Island (locality 15) and 23 it is the only haplotype found in this island. 24 For the CO1 gene, the most common haplotype (20 individuals) within haplogroup 2 is 25 restricted to the coastal Moroccan localities situated in-between the localities of Abtith 26 (locality 38) and Merja Zerga (locality 26). One haplotype (15 individuals) is endemic to the 27 Hoggar Mountains (localities 17 to 22) and Al Baten (locality 25), and it is the only haplotype 28 found in the Hoggar Mountains. One haplotype (4 individuals) is evenly distributed (present 29 in localities 25 and 26). 30 31 Negative values of Fu’s Fs were obtained for all genes and all haplogroups but they were only 32 significant for haplogroup 1 and the 16S gene, and for haplogroup 2 and the D-loop region 33 (Table 2). It was almost significant P = 0.03) for haplogroup 1a and the D-loop region.

1 Positive values of R2 were obtained for all genes and all haplogroups, but they were only 2 significant for haplogroup 1a and the D-loop region. 3 Visual examination of network characteristics (i.e. a star-like haplotype network where 4 many new haplotypes are derived from one dominant haplotype via single of few mutation 5 steps, indicating that they are the product of recent mutation events) indicate population 6 expansion in haplogroup 1a for the D-loop region, and in haplogroup 1 for the 16S and CO1 7 genes. The central haplotype in the network, which could be interpreted as ancestral, is widely 8 distributed and gives no information about the geographical origin of the population 9 expansion. 10 To identify the center of a population expansion it can be interesting to compare haplotype 11 and nucleotide diversity between populations. However this type of analysis cannot be done 12 here due to low sample size per population. 13 14 Our phylogenetic tree, based on 16S and D-loop sequences (Figure 4), shows that haplogroup 15 1 is monophyletic (posterior probability = 1.00). The monophyly of haplogroup 2 is not 16 confirmed (posterior probability = 0.69). All specimens from Algeria (in blue) cluster together 17 in the tree, as well as all specimens from Lampedusa Island (in purple). Specimens from 18 Tunisia (in red) cluster in two distinct groups, and the two specimens from Egypt (in green) 19 are phylogenetically nested within one of these groups. 20 21 Based on our BEAST analyses and the substitution rates of Stöck et al. (2008), the TMRCA 22 of all B. boulengeri is estimated at ca 0.773 Mya (range 0.210‒1.787 Mya) and 0.601 Mya 23 (0.128‒1.716) according to our D-loop and 16S results, respectively. Based the separation 24 between the Iberian and the African populations of Bufo bufo at ca 3.07 Ma, the TMRCA of 25 all B. boulengeri is estimated at ca 1.140 Mya (range 0.242‒2.455 Mya) with the 16S 26 sequences. 27 28 Species distribution modelling 29 To study the relationships between genetic diversity and possible glacial refugia, we built a 30 species distribution model (Figure 5) for B. boulengeri based on the known-presence 31 localities of this species and bioclimatic layers. The AUC value is 0.980, which is considered 32 to correspond to a useful predictive model. Under present-day climatic conditions the model 33 reveals high climatic habitat suitability for this species across most coastal Morocco and 34 Algeria (in-between the Atlantic Ocean or Mediterranean sea and the Atlas Mountains), from

1 the Souss valley to Alger, across coastal Tunisia, and across coastal Libya from the western 2 border to Misourata, and from Adjedabia to Derna. High climatic habitat suitability also exists 3 for this species south of the Atlas Mountains, at the limit with the Sahara desert, from Oued 4 Guir in Morocco to Biskra in Algeria. Finally high climatic habitat suitability is recorded in 5 the Hoggar Mountains. 6 According to the CCSM model, climatically suitable habitat for B. boulengeri was patchily 7 distributed during the LGM on the Atlantic coast of Morocco, on the coast at the Moroccan- 8 Algerian border, in coastal Tunisia, south of the Atlas Mountains, in the Hoggar Mountains, 9 in south-east Libya and in most of Egypt, except the coast and the south-east. The MIROC 10 model predicted larger areas of climatically suitable habitat than the CCSM model for B. 11 boulengeri during the LGM in the south, with an almost continuous high habitat suitability 12 range from the south of the Atlas Mountains to the Hoggar Mountains and in most Libya and 13 Egypt. 14 For the Mid-Holocene the CCSM and MIROC models give similar results. The distribution of 15 climatically suitable habitat for B. boulengeri at this period is predicted to be more or less 16 similar to present-day conditions, except lower habitat suitability in the Anti-Atlas, Atlas and 17 Middle Atlas mountains, in the Rif Mountains, and east of the Moulouya River in Morocco. 18 19 Discussion 20 Low phylogeographic structure across B. b. boulengeri geographic range 21 Our genetic results are mostly congruent with previous studies (Batista et al. 2006; Stöck et 22 al. 2006; Stöck et al. 2008) in showing low genetic diversity in B. b. boulengeri across North 23 Africa. However our more extensive sampling scheme, in terms of number of specimens and 24 localities, allowed us to highlight some mtDNA phylogeographic structure within this species: 25 one haplogroup (1) is endemic to coastal Morocco, from Aglou to Merja Zerga (Figure 3), 26 and another haplogroup (2) is widely distributed throughout North Africa, from coastal 27 Morocco to Egypt. Moreover, within haplogroup 2, several haplotypes are significantly 28 underdistributed. Low genetic diversity values and no phylogeographic structure were 29 obtained for the nDNA gene (RAG1). These results suggest that the mitochondrial 30 phylogeographic structure of the species has a recent origin. 31 This low mitochondrial structure at continental scale contrasts with the east-west divergence 32 that has been commonly found in previous phylogeographic studies of North African 33 organisms (see Nicolas et al. 2015 for a series of examples). The African green toad is a 34 relatively large amphibian that is supposed to travel considerable distances during its 4‒5 year

1 lifespan, like its congener B. viridis, allowing considerable gene flow between populations 2 (Langton 1990). Low genetic variation across North Africa was also observed for 3 Amietophrynus mauritanicus (Schlegel, 1841), another toad species with high vagility and 4 wide ecological niche preferences (Harris and Perera 2009). We thus hypothesise that a 5 combination of high vagility and continuously suitable habitats across North Africa across 6 glacial and interglacial periods (see Fig. 4) explains the unusual lack of mitochondrial 7 divergence between eastern and western North Africa in B. boulengeri. 8 Given the current distribution of the species (Figure 1) one could however expect to find 9 genetically divergent populations in the islands of Kerkennah and Lampedusa, the Hoggar 10 Mountains and the Sabha district. In agreement with this hypothesis we found haplotypes 11 endemic to Lampedusa Island (locality 15) and the Kerkennah Islands (locality 11) with the 12 mtDNA markers. However these haplotypes differ from mainland Africa haplotypes by few 13 mutations (1 mutation for the closest haplotype), and one haplotype was found in common in 14 both Kerkennah Islands and Egypt (locality 2). With the fastest evolving mtDNA gene we 15 found only one haplotype in the Hoggar Mountains, and it is endemic to these mountains. 16 This haplotype differ by two mutations from the closest haplotype (present in locality 6). 17 With the 16S and Co1 genes we found only one haplotype in the Hoggar Mountains, but this 18 haplotype was also present in the locality of Al Baten (locality 25). No haplotype endemic to 19 Sabha district was recorded in our analyses. Hence, no or little genetic differentiation was 20 observed for populations from the islands of Kerkennah and Lampedusa, the Hoggar 21 Mountains or the Sabha district. The islands of Kerkennah and Lampedusa are separated from 22 mainland Tunisia by low-depth waters (maximum depth around -70m according to Google 23 Earth) and were thus connected to mainland North Africa as recently as the LGM when sea 24 levels were around 120m lower than today (Rohling et al. 1998, Ray and Adams 2001). Lack 25 of divergence from mainland populations is thus not unexpected. Lampedusa and other 26 islands around Sicily have been colonized by one amphibian (B. boulengeri) and several 27 reptiles (see Corti et al. 1997), and phylogeographic studies suggest that a mix of natural 28 range expansions and human-mediated colonization events explain the current distribution of 29 amphibians and reptiles in these islands (Corti et al. 1997; Stöck et al. 2016). 30 31 Ecological niche modelling help understanding the observed phylogeographical pattern 32 Using different calibration priors resulted in slightly different divergence time estimates, but 33 our mtDNA data indicate that all intraspecific divergent events within the African Green toad 34 probably occurred during the last My. The Quaternary Period was dominated by Ice Ages and

1 involved repeated global cooling and increasing of the Arctic and Antarctic ice sheets (Hewitt 2 2004), and North Africa regularly shifted from wetter to drier climatic conditions (deMenocal 3 2004; Schuster et al. 2006; Tabel et al. 2016). In the absence of available climatic data for 4 oldest glacial periods, we used the LGM conditions as a proxy of the paleo-climatic 5 conditions of glacial periods. However it should be stressed that the intensity of glaciations 6 was not always the same from one glacial cycle to another and that even for the LGM several 7 uncertainties are associated with the climate data. One major point is the reliability of global 8 circulation models for the LGM climate (Schorr et al. 2012). In our data, uncertainty resulting 9 from climate model is evident from the LGM species distribution obtained with the CCSM 10 versus MIROC models (Figure 5). However, despite these drawbacks, results of our climatic 11 niche modelling analyses help to understand the observed phylogeographic pattern. 12 Our results for the LGM with the MIROC model suggest an expansion of the distribution 13 range of B. b. boulengeri during glacial periods. This might have facilitated the species’ 14 dispersal and subsequent gene ﬂow between most North African localities, explaining the 15 quasi absence of genetic distinctiveness of Hoggar Mountains or Sabha district populations 16 compared to other North African populations. 17 In Coastal Morocco no haplotype was found to be geographically underdistributed, which is 18 congruent with the high climatic habitat suitability modeled for the African green toad in this 19 area throughout the Quaternary. 20 We found one haplotype endemic to coastal Tunisia (localities 9, 11 and 12) and coastal 21 Libya (locality 6). This pattern is easily explain by the actual and mid-Holocene modeled 22 distributions of the species; highly climatically suitable habitat with no significant climatic 23 barrier being present in this area at these periods. 24 25 Origin of the two mtDNA haplogroups 26 Our mtDNA sequencing results identify two haplogroups within B. b. boulengeri: one 27 endemic to coastal Morocco, from Aglou to Merja Zerga (Figure 3), and another one widely 28 distributed throughout North Africa, from coastal Morocco to Egypt. Thus the geographical 29 range of the first haplogroup is embedded in the geographical range of the second haplogroup. 30 Two hypotheses can be proposed to explain this pattern: i) vicariance of an eastern and 31 western haplogroup followed by secondary expansion of the eastern haplogroup westward or 32 ii) retention of deep mtDNA diversity in Morocco followed by loss of diversity during 33 eastward expansion of the species out of Morocco.

1 i) Vicariance, probably around 1.1‒0.7 Ma, could be linked to the late Early Pleistocene 2 global climate transition, and particularly to MIS 22. MIS 22 (at ~0.87Ma) within the late 3 Early Pleistocene global climate transition represents the first prominent cold stage of the 4 Pleistocene. A burst of dust production and therefore aridity occurred at that time, and it 5 corresponds to a period of important vegetation changes (forest withdrawal and onset of 6 steppe or open vegetation), faunal turnover for mammals in North Africa and migration pulses 7 of large herbivores and hominins from North Africa and Eastern Europe into Southern 8 European refugia (Muttoni et al. 2010; Stoetzel 2013). At this period B. b. boulengeri may 9 have survived in two distinct allopatric areas, leading to allopatric diversification. The first 10 haplogroup would then have remained geographically restricted to the Atlantic Coast of 11 Morocco (between Aglou and the Merja Zerga), its dispersion eastward and southward being 12 limited by the Rif and Atlas Mountains, respectively. The second haplogroup would have 13 spread throughout large parts of North Africa as the result of broad climatic suitability for the 14 species South of the Atlas mountains and the Saharan Atlas during most of the Quaternary, 15 and particularly during glacial maxima (according to the MIROC model). 16 ii) Alternatively, deep genetic diversity could have persisted in Morocco thanks to the 17 continuous occurrence of climatically favorable conditions for the species and the complex 18 topography of the country. A recent population expansion eastward throughout large parts of 19 North Africa, favored by the broad climatic suitability for the species south of the Atlas 20 mountains and the Saharan Atlas during most of the Quaternary, would have resulted in a loss 21 of genetic diversity, with only the haplogroup 2 retained in the area of expansion. 22 Note that both scenarios imply a population expansion of haplogroup 2, for which we find no 23 evidence in the tests we performed (see results), but the sparse sampling of haplogroup 2 24 outside Morocco might limit the power of the tests. The topology of our phylogenetic tree, 25 based on 16S and D-loop sequences (Figure 4), supports the hypothesis of retention of deep 26 mtDNA diversity in Morocco followed by loss of diversity during eastward expansion of the 27 species out of Morocco. Haplogroup 2 does not form a strongly supported clade in our 28 phylogenetic tree. All specimens from Algeria cluster together in the tree, as well as all 29 specimens from Lampedusa Island. Specimens from Tunisia cluster in two distinct groups, 30 and the two specimens from Egypt are phylogenetically nested within one of these groups. 31 These results suggest that several waves of eastward expansion occurred. 32

1 To conclude, our integrative approach, combining phylogeography and climatic niche 2 modelling, allowed us to propose a plausible hypothesis explaining the low phylogeographic 3 structure observed within B. b. boulengeri across the Maghreb region. 4 5 Acknowledgments 6 This work has been funded by the ANR project PEX04-MOHMIE. Molecular analyses were 7 supported by the ‘Action Transversale du Museum: Taxonomie moléculaire, DNA Barcode & 8 Gestion Durable des Collections’, the Service de Systématique Moléculaire of the French 9 National Museum of Natural History (UMS 2700, Paris, France) and the network 10 ‘Bibliothèque du Vivant’ funded by the CNRS, the Muséum National d’Histoire Naturelle, 11 the INRA, and the CEA (Genoscope). DNA sampling in Morocco was permitted thanks to the 12 ‘Haut Commissariat aux Eaux et Forêts et à la Lutte Contre la Désertification (authorizations 13 15 HCEFLCD/DLCDPN/DPRN/CFF)’ in the frame of a joint collaboration between the 14 Institut Scientifique of Rabat and the Muséum National d’Histoire Naturelle. We are grateful 15 to the people who sent us tissue samples and/or field observations, particularly Abdelkhader 16 Allali, Benjamin Allegrini, Menad Beddek, Wouter Beukema, Pierre Beaubrun, Jacques 17 Bons, Fabrice Cuzin, Boualem Dellaoui, Rémi Destre, Bertrand Delprat, Guy Durand, 18 Mohamed El-Ghazrani, Pascal Escudié, Thomas Gallix, Michel Geniez, Yves Hingrat, Paul 19 Isenmann, Joan Martinez i Giner, Raphaël Leblois, José Antonio Mateo, Aïssa Moali, Olivier 20 Peyre, Juan-Manuel Pleguezuelos, Philippe Roux, Paul Soto, Michel Thévenot, Gilles 21 Trochard, Julien Viglione and Sohaib Liefried. Finally, we thank two anonymous reviewers 22 for their constructive remarks that allowed to improve this manuscript. 23 24 References 25 Akaike H (1973) Information theory as an extension of the maximum likelihood principle. In: 26 Petrov BN, Csaki F (eds), Second International Symposium on Information Theory. 27 Akademiai Kiado, Budapest, pp 267–281. 28 Alvarez Y, Mateo JA, Andreu AC, Diaz-Paniagua C, Diez A, Bautista JM (2000) 29 Mitochondrial DNA haplotyping of Testudo graeca on both continental sides of the 30 Straits of Gibraltar. J Hered 91:39-41. 31 Amor N, Farjallah S, Ben-Yacoub S, Merella P, Khaled S (2011) Morphological Variation of 32 the African Green Toad, Bufo boulengeri (Amphibia: Anura) in Tunisia. Pakistan J 33 Zool 43:921-926.

1 Aulagnier S, Thevenot M (1986) Catalogue des mammifères sauvages du Maroc. Trav Ins Sci 2 Sér Zool 41:1-163. 3 Baha El Din S (2006) A guide to the Reptiles and Amphibians of Egypt. American University 4 in Cairo Press, Cairo. 5 Bailon S (2000) Amphibiens et reptiles du Pliocène Terminal d’Ahl al Oughlam (Casablanca, 6 Maroc). Geodiversitas, 22:539-558. 7 Bandelt HJ, Forster P, Rohl A (1999) Median-joining networks for inferring intraspecific 8 phylogenies. Mol Biol Evol 16:37-48. 9 Barata M, Harris DJ, Castilho R (2008) Comparative phylogeography of northwest African 10 Natrix maura (Serpentes: Colubridae) inferred from mtDNA sequences. Afr Zool 11 43:1-7. 12 Batista V, Carranza S, Carretero MA, Harris DJ (2006a) Genetic variation within Bufo viridis: 13 evidence from mitochondrial 12S and 16S RRNA DNA sequences. Butll Soc Cat 14 Herp 17:24-33. 15 Ben Faleh A, Granjon L, Tatard C, Ben Othmen A, Said K, Cosson JF (2012) 16 Phylogeography of the Greater Egyptian Jerboa (Jaculus orientalis) (Rodentia: 17 Dipodidae) in Mediterranean North Africa. J Zool 286:208–220. 18 Ben Hassine J, Gutiérrez-Rodríguez J, Escoriza D, Martínez-Solano I (2016) Inferring the 19 roles of vicariance, climate and topography in population differentiation in 20 Salamandra algira (Caudata, Salamandridae). J Zool Syst Evol Res 54:116-126. 21 Beukema W, De Pous P, Donaire-Barroso D, Bogaerts S, Garcia-Porta J, Escoriza D, Arribas 22 OJ, El Mouden EH, Carranza S (2013) Review of the systematics, distribution, 23 biogeography and natural history of Moroccan amphibians. Zootaxa 3661:1-60. 24 Bons J, Geniez P (1996) Amphibiens et Reptiles du Maroc (Sahara Occidental compris), 25 Atlas biogéographique. Asociación Herpetológica Española, Barcelona, Spain. 26 Brown RP, Suarez NM, Pestano J (2002) The Atlas mountains as a biogeographical divide in 27 North-West Africa: evidence from mtDNA evolution in the Agamid lizard Agama 28 impalearis. Mol Phylogenet Evol 24:324-332. 29 Brunet P, Sanuy D, Oromi N, Ait Hammou M, Dahmani W (2009) Anuran studies from 30 Tiaret region, north-west of Algeria. Bol Asoc Herpetol Esp 20:68-72. 31 Chiari Y, Vences M, Vieites DR, Rabemananjara F, Bora P, Ramilijaona Ravoahangimalala 32 O, Meyer A (2004) New evidence for parallel evolution of colour patterns in 33 Malagasy poison frogs (Mantella). Mol Ecol 13:3763-3774.

1 Collins WD, Bitz CM, Blackmon ML, Bonan GB, Bretherton CS, Carton JA, Chang P, 2 Doney SC, Hack JJ, Henderson TB, Kiehl JT, Large WG, McKenna DS, Santer BD, 3 Smith RD (2006) The Community Climate System Model Version 3 (CCSM3). J 4 Climate 19:2122-2143. 5 Corti, C, Lo Cascio, P, Vanni, S, Turrisi, GF, Vaccaro, A (1997). Amphibians and reptiles of 6 the circumsicilian islands: new data and some considerations. Boll Mus reg Sci nat 7 Torino 15: 179-211. 8 Damas-Moreira I, Tome B, Harris DJ, Maia JP, Salvi D (2014) Moroccan herpetofauna: 9 distribution updates. Herpetozoa 27:96-102. 10 deMenocal P (2004) African climate change and faunal evolution during the Pliocene- 11 Pleistocene. Earth Planet Sc Lett 220:3-24. 12 Donaire D, Beukema W, de Pous P, Raul del Canto G (2011) A distributional review of Bufo 13 boulengeri Lataste, 1879 in northern Morocco with emphasis on occurrence in the Rif 14 Mountains. Herpetology Notes 3:71-74. 15 Douady CJ, Catzeflis F, Raman J, Springer MS, Stanhope MJ (2003) The Sahara as a 16 vicariant agent, and the role of Miocene climatic events, in the diversification of the 17 mammalian order Macroscelidea (elephant shrews). Proc Natl Acad Sci U S A 18 100:8325-8330. 19 Drake NA, Blench RM, Armitage SJ, Bristow CS, White KH (2011) Ancient watercourses 20 and biogeography of the Sahara explain the peopling of the desert. Proc Natl Acad Sci 21 U S A 108:458-462. 22 Drummond AJ, Rambaut A (2007) BEAST: Bayesian evolutionary analysis by sampling 23 trees. BMC Evol Biol 7:214. 24 Dubois A, Crochet PA, Dickinson EC, Nemésio A, Aescht E, Bauer AM, Blagoderov V, Bour 25 R, De Carvalho MR, Desutter-Grandcolas L, Frétey T, Jäger P, Koyamba V, Lavilla 26 EO, Löbl I, Louchart A, Malécot V, Schatz H, A. O (2013) Nomenclatural and 27 taxonomic problems related to the electronic publication of new nomina and 28 nomenclatural acts in zoology, with brief comments on optical discs and on the 29 situation in botany. Zootaxa 3735:1-94. 30 Dufresnes C, Bonato L, Novarini N, Betto-Colliard C, Perrin N, Stöck M (2014) Inferring the 31 degree of incipient speciation in secondary contact zones of closely related lineages of 32 Palearctic green toads (Bufo viridis subgroup). Heredity 113:9-20. 33 Elith J, Graham CH, Anderson RP, Dudik M, Ferrier S, Guisan A, Hijmans RJ, Huettmann F, 34 Leathwick JR, Lehmann A, Li J, Lohmann LG, Loiselle BA, Manion G, Moritz C,

1 Nakamura M, Nakazawa Y, Overton JM, Peterson T, Phillips SJ, Richardson K, 2 Scachetti-Pereira R, Schapire RE, Soberón J, Williams S, Wisz MS, Zimmermann NE 3 (2006) Novel methods improve prediction of species' distributions from occurrence 4 data. Ecography 29:129–151. 5 Elith J, Kearney M, Phillips S (2010) The art of modelling range-shifting species. Meth Ecol 6 Evol 1:330-342. 7 Escoriza D, Comas MM, Donaire D, Carranza S (2006) Rediscovery of Salamandra algira 8 Bedriaga, 1883 from the Beni Snassen massif (Morocco) and phylogenetic 9 relationships of North African Salamandra. Amphibia-Reptilia 27:448-455. 10 Essghaier MFA, Taboni IM, Etayeb KS (2015) The diversity of wild animals at Fezzan 11 Province (Libya). Biodiversity Journal 6:245-252. 12 Fahd S, Mediani M (2007) Herpetofaune du bassin versant de Oued Laou. Report by 13 Université de Tétouan. 14 Filippi E, Chetoui A, D’Alterio GL, Parker G, Woodfine T (2011) Bufo boulengeri: field 15 observations on breeding phenology at a new Saharan site in Southern Tunisia. 16 Herpetology Notes 4:215-217. 17 Flot JF (2010) Seqphase: a web tool for interconverting phase input/output files and fasta 18 sequence alignments. Mol Ecol Res 10:162-166. 19 Fourcade Y, Engler JO, Rodder D, Secondi J (2014) Mapping species distributions with 20 MAXENT using a geographically biased sample of presence data: a performance 21 assessment of methods for correcting sampling bias. PLoS One 9:e97122. 22 Fritz U, Barata M, Busack SD, Fritzsch G, Castilho R (2006) Impact of mountain chains, sea 23 straits and peripheral populations on genetic and taxonomic structure of a freshwater 24 turtle, Mauremys leprosa (Reptilia, Testudines, Geoemydidae). Zool Scripta 35:97– 25 108. 26 Frynta D, Kratochvil L, Moravec J, Benda P, Dandova R, Kaftan M, Klosova K, Mikulova P, 27 Nova P, Schwarzova L (2000) Amphibians and reptiles recently recorded in Libya. 28 Acta Soc Zool Bohem 64:17-26. 29 Fu YX (1997) Statistical tests of neutrality of mutations against population growth, 30 hitchhiking and background selection. Genetics 147:915-925. 31 García-Muñoz E, Jorge F, Rato C, Carretero MA (2010) Four types of malformations in a 32 population of Bufo boulengeri (Amphibia, Anura, Bufonidae) from the Jbilet 33 Mountains (Marrakech, Morocco). Herpetology Notes 3:267-270.

1 Goebel AM, Donnelly JM, Atz ME (1999) PCR primers and amplification methods for 12S 2 ribosomal DNA, the control region, cytochrome oxidase I, and cytochrome b in 3 bufonids and other frogs, and an overview of PCR Primers which have amplified 4 DNA in amphibians successfully. Mol Phylogenet Evol 11:163-199. 5 Harris DJ, Perera A (2009) Phylogeography and genetic relationships of North African Bufo 6 mauritanicus Schlegel, 1841 estimated from mitochondrial DNA sequences. Biologia 7 64:356-360. 8 Harris DJ, Carretero MA, Brito JC, Kaliontzopoulou A, Pinho C, Perera A, Vasconcelos R, 9 Barata M, Barbosa D, Carvalho S, Fonseca M, Perez-Lanuza G, Rato C (2008) Data 10 on the distribution of the terrestrial herpetofauna of Morocco: records from 2001- 11 2006. Herpetological Bulletin 103:19-28. 12 Hasumi H, Emori S, eds. (2004) K-1 Coupled GCM (MIROC) Description. K-1 Model 13 developers tech Rep 1, Center for Climate System Research, University of Tokyo, 14 Tokyo, Japan. 15 Hewitt GM (2004) Genetic consequences of climatic oscillations in the Quaternary. Philos 16 Trans R Soc B 359:183-195. 17 Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution 18 interpolated climate surfaces for global land areas. International Journal of 19 Climatology 25:1965-1978. 20 Ho SY, Duchene S (2014) Molecular-clock methods for estimating evolutionary rates and 21 timescales. Molecular Ecology 23:5947-5965. 22 Ibrahim A (2008) Contribution to the herpetology of southern Libya. Acta Herpetologica 23 3:35-49. 24 Ibrahim A (2013) Amphibian in Libya: a status report. Basic and Applied Herpetology 25 27:101-106. 26 ICZN (2014) Zoological Nomenclature and Electronic Publication—a reply to Dubois et al. 27 (2013). Zootaxa 3779:3-5. 28 Kelly RP, Oliver TA, Sivasundar A, Palumbi SR (2010) A method for detecting population 29 genetic structure in diverse, high gene-flow species. J Hered 101:423-436. 30 Langton TES (1990) Amphibians and roads. Proceedings of the Conference held in 31 Proceedings of the Toad Tunnel Conference Rendsburg, Federal Republic of 32 Germany. 33 Le Houérou HN (1997) Climate, flora and fauna changes in the Sahara over the past 500 34 million years. J Arid Environ 37:619-647.

1 Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA 2 polymorphism data. Bioinformatics 25:1451-1452. 3 Márquez-Rodríguez J (2014) Bufotes boulengeri (African Green Toad): New reproductive 4 population in Tunisia. Herpetological Bull 127:33-34. 5 Martínez-Solano I, Sindaco R, Romano A (2015) Bufotes boulengeri. The IUCN Red List of 6 Threatened Species 2015: eT153568A74497730 7 http://dxdoiorg/102305/IUCNUK2015-1RLTST153568A74497730en. 8 Muttoni G, Scardia G, Kent DV (2010) Human migration into Europe during the late Early 9 Pleistocene climate transition. Palaeogeogr, Palaeoclimat, Palaeoecol 296:79-93. 10 Nei M (1987) Molecular Evolutionary Genetics. Columbia University Press, New York. 11 Nicolas V, Mataame A, Crochet PA, Geniez P, Ohler A (2015) Phylogeographic patterns in 12 North African water frog Pelophylax saharicus (Anura: Ranidae). J Zool Syst Evol 13 Res 53:239-248. 14 Nicolas V, Ndiaye A, Benazzou T, Souttou K, Delapre A, Denys C (2014) Phylogeography of 15 the North African dipodil (Rodentia: Muridae) based on cytochrome-b sequences. J 16 Mammal 95:241-253. 17 Nylander JA (2004) MrModeltest v2. Program distributed by the author. Evolutionary 18 Biology Centre, Uppsala University, Uppsala, Sweden. 19 Palumbi SR, Martin A, Romano S, McMillan WO, Stice L, Grabowski G (1991) The Simple 20 Fool’s Guide to PCR. University of Hawaii Press, Honolulu. 21 Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modelling of species 22 geographic distributions. Ecol Modelling 190:231-239. 23 Posada D, Crandall K (2001a) A comparison of different strategies for selecting models of 24 DNA substitution. Systematic Biology 50:580–601. 25 Posada D, Crandall KA (2001b) Intraspecific gene genealogies: trees grafting into networks. 26 Trends Ecol Evol 16:37-45. 27 Radosavljevic A, Anderson RP (2014) Making better MAXENT models of species 28 distributions: complexity, overfitting and evaluation. J Biogeogr 41:629-643. 29 Ramos-Onsins SE, Rozas J (2002) Statistical properties of new neutrality tests against 30 population growth. Mol Biol Evol 19:2092-2100. 31 Ray N, Adams JM (2001) A GIS-based Vegetation Map of the World at the Last Glacial 32 Maximum (25,000-15,000 BP). Internet Archaeology 11.

1 Recuero E, Iraola A, Rubio X, Machordom A, Garcia-Paris M (2007) Mitochondrial 2 differentiation and biogeography of Hyla meridionalis (Anura : Hylidae): an unusual 3 phylogeographical pattern. J Biogeogr 34:1207-1219. 4 Rognon PP (1989) Biogéographie d'un désert. Plon, Paris. 5 Rohling EJ, Fenton M, Jorissen FJ, Bertrand P, Ganssen G, Caulet JP (1998) Magnitudes of 6 sea level lowstands of the past 500,000 years. Nature 394: 162-165. 7 Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed 8 models. Bioinformatics 19:1572–1574. 9 Rosado D, Harris DJ, Perera A, Jorge F, Tome B, Damas-Moreira I, Tavares I, Estrela H, 10 Sousa A, Pereira A, Mantovani M, Salvi D (2016) Moroccan herpetofauna distribution 11 updates including a DNA barcoding approach. Herpetozoa 28:171–178. 12 San Mauro D, Gower DJ, Oommen OV, Wilkinson M, Zardoya R (2004) Phylogeny of 13 caecilian amphibians (Gymnophiona) based on complete mitochondrial genomes and 14 nuclear RAG1. Mol Phylogenet Evol 33:413-427. 15 Schorr G, Holstein N, Pearman PB, Guisan A, Kadereit JW (2012) Integrating species 16 distribution models (SDMs) and phylogeography for two species of Alpine Primula. 17 Ecol Evol 2:1260-1277. 18 Schuster M, Duringer P, Ghienne JF, Vignaud P, Mackaye HT, Likius A, Brunet M (2006) 19 The age of the Sahara desert. Science 311:821. 20 Sicilia A, Marrone F, Sindaco R, Turki S, Arculeo M (2009) Contribution to the knowledge 21 of Tunisian amphibians: notes on distribution, habitat features and breeding 22 phenology. Herpetology Notes 2:107-132. 23 Speybroeck J, Beukema W, Bok B, van der Voort J, Velikov I (2016) Field Guide to the 24 Amphibians and Reptiles of Britain and Europe. Bloomsbury, London & New York. 25 Stephens M, Scheet P (2005) Accounting for decay of linkage disequilibrium in haplotype 26 inference and missing data imputation. Am J Hum Genet 76:449-462. 27 Stephens M, Smith NJ, Donnelly P (2001) A new statistical method for haplotype 28 reconstruction from population data. Am J Hum Genet 68:978-989. 29 Stöck, M, Grifoni, G, Armor, N, Scheidt, U, Sicilia, A, Novarini, N (2016). On the origin of 30 the recent herpetofauna of Sicily: Comparative phylogeography using homologous 31 mitochondrial and nuclear genes. Zoologischer Anzeiger 261: 70-81. 32 Stöck M, Moritz C, Hickerson M, Frynta D, Dujsebayeva T, Eremchenko V, Macey JR, 33 Papenfuss TJ, Wake DB (2006) Evolution of mitochondrial relationships and

1 biogeography of Palearctic green toads (Bufo viridis subgroup) with insights in their 2 genomic plasticity. Mol Phylogenet Evol 41:663-689. 3 Stöck M, Sicilia A, Belfiore NM, Buckley D, Lo Brutto S, Lo Valvo M, Arculeo M (2008) 4 Post-Messinian evolutionary relationships across the Sicilian channel: mitochondrial 5 and nuclear markers link a new green toad from Sicily to African relatives. BMC Evol 6 Biol 8:56-56. 7 Stoetzel E (2013) Late Cenozoic micromammal biochronology of northwestern Africa. 8 Palaeogeogr Palaeoclimatol Palaeoecol 392:359–381. 9 Tabel J, Khater C, Rhoujjati A, Dezileau L, Bouimetarhan I, Carre M, Vidal L, Benkaddour 10 A, Nourelbait M, Cheddadi R (2016) Environmental changes over the past 25 000 11 years in the southern Middle Atlas, Morocco. J Quater Sc 31:93-102. 12 Varela S, Lima-Ribeiro MS, Terribile LC (2015) A short guide to the climatic variables of the 13 Last Glacial Maximum for biogeographers. PLoS One 10:e0129037. 14 Vences M, de Pous P, Nicolas V, Díaz-Rodríguez J, Donaire D, Hugemann K, Hauswaldt JS, 15 Amat F, Barnestein JAM, Bogaerts S, Bouazza A, Carranza S, Galán P, de la Vega 16 JPG, Joger U, ansari A, El Mouden EH, Ohler A, Sanuy D, Slimani T, Tejedo M 17 (2014) New insights on phylogeography and distribution of painted frogs 18 (Discoglossus) in northern Africa and the Iberian Peninsula. Amphibia Reptilia 19 35:305-320. 20 Warren DL, Glor RE, Turelli M (2010) ENMTools: a toolbox for comparative studies of 21 environmental niche models. Ecography 33:607-611. 22 23 24

1 Figure legends 2 Figure 1: Map showing the geographical range (shaded area) of Bufotes boulengeri 3 boulengeri according to Martínez-Solano et al. (2015), sampling localities (black dots), and 4 the main mountain ranges (light grey) and districts cited in the text.

5 6

1 Figure 2. Median-joining network of D-loop (A), 16S (B), CO1 (C) and RAG1 (D) 2 haplotypes of Bufotes boulengeri boulengeri. Circle sizes are proportional to the number of 3 similar haplotypes observed in the data set. The number of mutations between haplotypes is 4 proportional to the length of branches. For the D-loop, 16S and CO1 DNA fragments two 5 haplogroups can be identified: haplogroup 1 (haplotypes in white), and haplogroup 2 6 (haplotypes in black). For the RAG1 fragment no genetic structure can be identified, but 7 haplotypes are coloured in white and black according to the mitochondrial results.

8 9

1 Figure 3. Geographical distribution of the main haplogroups obtained in the median-joining 2 network analyses of the D-loop (A), 16S (B) and CO1 (C) haplotypes of Bufotes boulengeri 3 boulengeri. White = only specimens belonging to haplogroup 1 are present in this locality; 4 black = only specimens belonging to haplogroup 2 are present in this locality; grey = 5 specimens of both haplogroups 1 and 2 co-occur in this locality.

1 Figure 4. Molecular phylogenetic tree inferred using Bayesian inference for the concatenated 2 data set (D-loop and 16S). The average Bayesian posterior probabilities (left) of two runs (pp 3 > 0.85) and bootstrap values (> 70%, right) calculated for 1,000 bootstrap replicates are 4 indicated for each node. Colours indicate the geographical origin of the samples: black = 5 Morocco, red = Tunisia, blue = Algeria, green = Egypt, purple = Lampedusa island.

6 7

1 Figure 5. Species distribution modelling of Bufotes boulengeri boulengeri as estimated by 2 Maxent for present-day conditions (current), for the last glacial maximum (LGM) and for the 3 mid-holocene. For paleoclimatic data estimates are based on the Model for Interdisciplinary 4 Research on Climate (MIROC) and on the Community Climate System Model (CCSM) 5 paleoclimatic models. Warmer colours show areas with higher probability of presence.

6 7

1 Supporting Information 2 Table S1. List of specimens used in this study, with museum numbers, geographic origins 3 (specimens are order by continent (Africa, Europe), country, latitude and longitude), 4 GenBank accession numbers (GB)haplogroup (L) and haplotype numbers (H) for each gene. 5 In bold: newly obtained sequence data. 6 7 Table S2. Primer sequences used in this study. 8 9 Table S3. Distribution records of African green toads (sorted by country name, latitude and 10 longitude) assembled from literature, museum collections (Global Biodiversity Information 11 Facility, http://www.gbif.org, last access date: 26 August 2016) and our own unpublished 12 databases. 13 14 Figure S1. Maps of localities used to train the Maxent species distribution model (SDM) for 15 Bufotes boulengeri boulengeri. 16 17 18

1 Table 1: Collection localities of Bufotes boulengeri boulengeri specimens included in this 2 study, with the number of specimens sequenced per DNA region and per locality. The symbol 3 “*” indicates that several collecting sites were sampled within a given locality (see 4 Supporting information Table S1 for details on collecting sites), but they were grouped in our 5 analyses for statistical reasons. Numbers in parentheses for the RAG1 gene indicate 6 specimens that were sequenced but not included in the final analyses because phasing was 7 doubtful. Localities are order by continent (Africa, Europe), country, latitude and longitude.

Country Locality Locality Latitude Longitude D- 16S CO1 RAG1 number loop on Figure 1 Italy Lampedusa island 15 35.508 12.600 6 5 - - Morocco Saqiyat al-Hamra, 4 km WWWNW Edchera 39 27.030 -13.093 1 1 1 1 15 km NNE Abtih 38 28.001 -11.367 1 1 1 1 7 km W Tantan 40 28.438 -11.174 1 1 1 1 (1) Aglou-Plage, oued Oudoudou 37 29.807 -9.826 1 1 1 - Ait Baha 36 30.130 -9.080 1 - - - Sidi Mokhtar* 35 31.579 -9.019 15 16 15 7 Sidi Chiker* 33 31.751 -8.736 9 9 9 4 (1) Oued N'Fis, 20 km W Marrakesh 34 31.630 -8.248 2 2 2 1 Douar Iben Brahim 32 32.220 -7.970 1 1 1 - Doulad Bouziri 31 32.658 -7.779 2 2 2 - Settat, Douar Setoute 30 33.001 -7.741 1 1 1 - 16 and 17 km EEENE Berrechid 28 33.296 -7.392 14 14 14 - Ouled Boughadi 29 33.076 -6.728 4 5 4 3 Merja Zerga* 26 34.818 -6.291 9 11 9 4 Dayet Er-Roumi* 27 33.752 -6.185 5 5 5 5 (1) High Atlas (ca 9.5 km SE Tounfite) 24 32.427 -5.156 2 - - - Dayet Termerzigat, near Merzouga 23 31.081 -4.042 - 1 1 - Al Baten 25 33.186 -3.990 6 7 6 5 Algeria Ghardaia 16 32.483 3.667 1 - - - Hoggar, locality 1 17 23.454 5.995 2 2 2 2 Hoggar, locality 2 18 23.126 5.989 3 4 3 -

Hoggar, locality 3 20 23.157 5.423 1 2 1 - Hoggar, locality 4 21 23.421 5.762 1 2 3 - Hoggar, locality 5 22 23.478 5.810 1 2 2 - Hoggar, locality 6 19 22.786 5.814 2 2 2 - Tunisia El Kef 13 36.166 8.700 2 2 - - Cap Bon, Lebna 12 36.728 10.931 3 3 - - Kerkennah Islands, Chergui* 11 34.820 11.258 5 5 - - Djerba island 10 33.800 10.900 1 - - - Nefta oasis 14 33.917 7.883 1 - - - Tataouine 9 32.903 10.416 3 3 - - Libya Al' Fiayi 8 26.533 13.317 1 - - - Gabroon lake 7 26.803 13.535 3 - - - Sahahhat 6 32.817 21.867 3 - - - Egypt Matrouh 5 28.750 26.980 2 - - - Oasis Dakhla 1 25.553 28.948 2 - - - 70 km S Alexandria 4 30.584 29.963 1 - - - Alexandria, El Menoufia 3 30.500 31.000 1 - - - Wadi El Natrun 2 30.413 30.285 2 2 2 - Unknown Unknown - 1 - - Total 122 113 88 35 (3) 1 2 3

1 Table 2: Diversity and neutrality estimates for the main haplogroups identified in the median- 2 joining analyses. Significant values are in bold. Number of sites (bp), number of sites 3 excluding indels (bpi), number of sequences (N), number of polymorphic sites (Np), number 4 of distinct haplotypes (Nh), haplotype diversity (Hd), nucleotide diversity (Pi) and average 5 number of nucleotide differences (k) are given.

bp bpi N Np Nh Hd Pi k Fu's Fs R2 D-loop All 545 534 122 40 34 0.930 ± 0.012 0.01498 ± 0.00063 8.000 Haplogroup 1a 545 542 33 15 10 0.744 ± 0.072 0.00326 ± 0.00061 1.769 -3.466 0.058 Haplogroup 1b 545 543 11 2 3 0.473 ± 0.162 0.00094 ± 0.00036 0.509 -0.659 0.165 Haplogroup 2 545 534 78 21 23 0.891 ± 0.025 0.00631 ± 0.00045 3.368 -8.138 0.078

16S All 586 583 113 14 12 0.791 ± 0.021 0.00617 ± 0.00019 3.597 Haplogroup 1 586 585 47 6 7 0.601 ± 0.054 0.00122 ± 0.00017 0.712 -3.316 0.065 Haplogroup 2 586 583 66 4 5 0.585 ± 0.051 0.00138 ± 0.00015 0.807 -0.72 0.103

CO1 All 806 806 88 23 15 0.833 ± 0.020 0.00629 ± 0.00018 5.071 Haplogroup 1 806 806 43 8 7 0.627 ± 0.063 0.00115 ± 0.00022 0.930 -2.442 0.065 Haplogroup 2 806 806 45 9 8 0.695 ± 0.047 0.00144 ± 0.00023 1.164 -1.642 0.074

RAG All 937 937 64 7 13 0.728 ± 0.042 0.00117 ± 0.00013 1.101 6 7 8